top of page
A deep learning solution for Geosynchronous Earth Orbit resident space object detection using imagery from RMIT’s Robotic Optical Observatory
Presented by:
Sai Krishna Reddy Vallapureddy
Sai Krishna Reddy Vallapureddy
RMIT University
Rasit Abay
FuturifAI
Brett A. Carter
RMIT University
Monica Wachowicz
RMIT University
Debaditya Acharya
RMIT University
Gail Iles
RMIT University
Suelynn Choy
RMIT University
"Space Situational Awareness (SSA) knowledge is crucial for the safe operation of space activities.
One of the challenges in SSA is the increasing number of objects that require continuous detection and
tracking. Fully automatic optical systems that leverage Machine Learning (ML) models to detect and
track Resident Space Objects (RSOs) will help address this challenge. Such systems would also enhance
catalogue maintenance and manoeuvre detection capabilities. Leveraging the recent advancements in
ML, a state-of-the-art model is proposed that uses the data from RMIT University’s Robotic Optical
Observatory (ROO) for detecting RSOs in geosynchronous orbit. The proposed ML model uses Feature
Pyramid Network (FPN) - a deep feature extractor that takes a single-scale image of arbitrary size as
input, and outputs proportionally sized feature maps at multiple levels. In this study, the ML model’s
ability to identify RSOs in ROO’s imagery is tested and compared against ROO’s existing RSO identification
software. It is intended that the ML model will be incorporated into ROO’s real-time observation
campaigns, and that the SSA data collected will be made publicly available to the SSA community."
Category:
Computing
bottom of page